3.3 Electrochemical characterization
3.3.2 Electrochemical impedance spectroscopy under controlled atmosphere 51
The electrochemical impedance spectroscopy experiments where carried out in a temperature range between 30−550
◦
C under controlled atmosphere with oxygen partial pressures be-tween10
−6−1 baras well as water partial pressures between10
−5−4.2·10
−3bar. These water partial pressures correspond to a range of 0−100 %relative humidity in the gas feed of the gas tight oven, where the experiments were conducted in.
The ovens used during for the experiments conducted over a period of three years in this thesis varied in size and gas volume, but were all equipped with a similar set of sensors and gas feed technology. A schematic drawing of the most sophisticated, custom build measure-ment setup is shown in figure 3.1. The experimeasure-mentally most suitable oven was the one with the smallest quartz tube volume. The tube was closed on one side (in the other ovens used, the tube was open on both ends), and the gas inlet was connected to an alumina capillary tube, that reached almost to the end of the tube in order to assure a quick gas atmosphere exchange controlled with mass flow controllers. The gases used in the feed were dry 5.0 O2 as well as dry 5.0 Ar from compressed gas cylinders both supplied by Praxair. The desired gas mixture was roughly adjusted by the corresponding flow ratio with gas flow controllers.
For humidified gas mixtures, the gas was humidified by directing it through an impinger in
Figure 3.1: Scheme of the measurement setup. The sample was heated within the oven. A Pt/PtRh thermocouple was used to measure the sample temperature together with an Keithley multimeter. The Novocontrol Alpha AN frequency analyzer was used for the impedance measurements. The gas atmosphere was controlled using mks mass-flow controllers and ZIROX SGM5EL oxygen electrolysis devices, which were also used to determine the oxygen partial pressure in the ovens gas outlet. The humidity was measured using a HYT 939. The temperature and atmosphere control, as well as the data logging of all senors and measurement equipment was done by a custom made LabView program, except for the Novocontrol impedance analyzer, which was controlled by the suppliers software.
which deionized water was kept at a constant temperature of 21
◦
C (±1°C). In order to measure the humidity of the gas atmosphere a humidity sensor was installed in the exhaust stream of the oven. This Hygrosens HYT 939 capacitive polymer based sensor is designed to measure the relative humidity in a range between 0−100 % (±1.8 %) at 23°C with an reproducibility of ±0.2 %.139,140 Different water partial pressures were adjusted by mixing humidified and dry gas streams, using computer controlled mass flow controllers. From the relative humidity measured with the sensor the respective water partial pressure in the gas mixture entering the oven was calculated.
The mks mass-flow controllers were used to adjust the desired oxygen partial pressure roughly by mixing the respective amount of oxygen with argon as carrier gas, using a volume flow of 50 sccm in total. The precise adjustment was done with a zirconia based oxygen electrolysis device SGM5EL from ZIROX. It served for removal or dosing of oxygen into the carrier gas. To an electrolysis cell, consisting of a zirconia tube with porous platinum electrodes, a potentiometric cell voltage is applied, keeping the desired set point of a po-tentiometric measurement cell constant by using a PID control circuit. As the current of the electrolysis cell is proportional to the oxygen pumped through the zirconia into or out
Figure 3.2: Scheme for the determination of the cell constant of an interdigital electrode on a thin film by equation (45). The ratio of the electrode distancedand the areaA(orange) define the cell constant. Ais a virtual electrode surface approximated by the finger lengthltimes the film thicknessh.
of the carrier gas, the oxygen partial pressure could be precisely set. Additionally, it was also possible to use the devices measurement cell only for monitoring the oxygen partial pressure within an relative error of <5 % in the exhaust stream of the oven as well.141 It was observed that there was no significant difference in oxygen partial pressure at the inlet and outlet of the oven, even for oxygen partial pressures as low as 10
−20bar, proving that the whole system was gas tight, as any leakage would have been detected that way.
The quartz glass tube, containing the sample and its connecting platinum wires, was mounted in a computer controlled tube furnace, in order to automatically adjust the oven/ sample temperature, which was measured by a thermocouple in the direct vicinity of the sample under test.
Within this tube the sample was contacted with 100µm thick platinum wires, that were mechanically fixed on the sample with an alumina based ceramic paste (Ceramabond 503 or 668).142 The electrical contact to the microelectrode was established by applying platinum paste (Heraeus) to the wire and the contact pad, prepared together with the interdigital electrode. The other end of the thin platinum wire was point welded to a thicker platinum wire, that was connected to the impedance analyzer. Figure 3.3 shows an image of the contacted sample with its structured electrode.
Each electrode had20 fingers of 3 mm length with a width of40µm and a gap of 40µm in between the fingers. Thus, the overall active electrode length l was derived from the number of intact electrode finger pairs (intact finger on both sides of the electrode) times their length. For a fully intact interdigital electrode a total electrode length l of 0.6 m was determined. This length times the film thickness h gives a reasonable first order approxi-mation for the electrode area A, needed to derive the cell constant (see figure 4.14). Here one assumes that the entire film beneath the electrode is active. Together with the distance between the electrode fingers d, the cell constant Z for each sample could be determined
Figure 3.3: Photograph of a contacted ceria thin film as prepared for the electrochemical impedance spec-troscopy. The photograph shows the substrate of 1x1 cm
2 covered with a 8CZO thin film, contacted by a platinum interdigital electrode. Each finger of the electrode is3 mmlong and40µmin width. In addition one can see the platinum wire that is used to establish the contact to the impedance analyzer, together with the platinum paste drops to ensure good electric contact and the ceramic paste drops ensuring mechanic stability.
individually by
Z =d/A= d
h·l. (45)
The resulting cell constants in this work were in the range of 19−158 cm
−1, depending on the number of intact fingers after the lift off process and the thin film thickness.
In order to characterize the samples with electrochemical impedance spectroscopy, typically a controlled atmosphere regarding oxygen and water partial pressure was adjusted while heating up the oven and sample to 550
◦C. After an equilibration time of about 4 h, at the end of which the impedance data did not change significantly anymore the temperature was lowered by 3
◦
C/min in 25
◦
C steps. After the next temperature level was reached, a temperature dwell was set for 3 h, before the ramp down to the next lower temperature level was started. At 300
◦
C typically the temperature steps were increased to50
◦
C per step, as here for most of the samples and atmosphere conditions the impedance was so large, that no reliable data could be measured. For temperatures around 150
◦
C and lower, the step size was decreased again to 25
◦
C per step, as here the impedance was smaller again for elevated humidities. It should be noted, that the dwell time was optimized regarding time efficiency from experiment to experiment. For higher temperatures the impedance data often stabilized after a short time already whereas this took longer at lower temperatures. Taking this into account the dwell times for the different temperature steps were set individually, in order to get reproducible data in a time efficient way. Without this a single impedance measurement for one atmospheric condition and the whole temperature range easily took about five days. Optimizing the dwell times, it was possible to reduce this to about 26 h, including the heating and atmosphere exchange and equilibration between two experiments.
All AC conductivity impedance measurements were carried out with a 2-electrode-4-wire setup connecting the sample to a Novocontrol alpha-A impedance analyzer, equipped with a ZG4 interface. In order to minimize the effect of leakage currents the Faraday cage was kept at the potential of the high impedance terminal, by connecting it to the driven shield of the ZG4 interface. The impedance spectra were recorded within a frequency range of10 mHzto 1 MHz with an oscillation voltage amplitude of 20 mV. The impedance data was exported from the Novocontrol software into ASCII files. Those were then combined with ASCII based data sets including temperature, partial pressure and humidity sensors of the oven setup using their common time stamp. A custom made plugin kindly provided by the vendor of the data analysis software RelaxIS (rhd-instruments) used in Version 2 and 3 allowed importing the impedance data together with the oven and gas atmosphere information.143 This turned out as a very reliable and convenient way to make sure that the environmental conditions are precisely known for any single data point of the impedance spectrum.ii) This is not the case for other typical setups, as the environmental information during the measurement are averaged over the impedance cycle. In addition the RelaxIS software package allows to handle large data sets of impedance measurement series efficiently. The results of the impedance data fitting were then exported to Origin 2018 (OriginLab corporation) for further data analysis and plotting purposes.144
ii)In the beginning of this project the data were then analyzed with the ZSimpWin software package. Unfor-tunately version 3.21, which was available, is not the latest version, suffering from compatibility problems with modern computer hard- and software. With this software the environmental parameters, like oven temperature, and gas partial pressures had to be correlated with the impedance data manually.
4 Results and discussion
In this chapter the results of the experiments are presented and discussed. In the first section the results of the target preparation are going to be outlined. It will be shown that the target preparation route chosen in this study was feasible for the preparation of such targets needed for the pulsed laser deposition process regarding the composition and their density. Therefore, the results of SEM, XRD and XPS measurements on the targets will be presented. A good characterization of the starting material for the PLD process is essential for a sufficient characterization of the thin films prepared from those targets. In a further section it will be shown that it was possible to prepare thin films with different morphologies via PLD and post-ablation annealing. Here, the focus will be set on the structure and morphology measurements by XRD, GI-XRD and SEM as well as EDS and XPS compositional analysis.
Finally, in the last section of this chapter the results of the EIS measurements are presented.
4.1 Morphology and chemical composition of the PLD targets
Characterizing the phase composition of a solid bulk material, one of the most important techniques nowadays is X-ray diffraction. Figure 4.1 shows the results of the XRD measure-ments on the 10CZO, 8CZO and 6CZO targets. The stacked graphs on the left show the XRD pattern as measured with the powder diffractometer in θ/2θ geometry between 20
◦
and 80
◦. The respective reflex assignments are based on the space group for the pattern given in the right graphs. In those plots the region between 69
◦ and 72
◦ is shown in detail for the three samples, demonstrating the splitting of the (004) reflex for the cubic space group into the (004) and (220)reflex for the tetragonal pattern. The analysis of the X-ray diffraction pattern show no evidence for any other phase than the one that was intended to be prepared by the route described in section 3.1.1. No impurity phases were detected.
Whereas the CeO2(10CZO) and the Ce0.8Zr0.2O2 (8CZO) target show a pattern that corre-sponds to the cubic space groupFm¯3m, the pattern of the Ce0.6Zr0.4O2 (6CZO) is assigned to the tetragonal space group P42/nmc.145–147
As described in section 2.1.2 tetragonality occurs when the cubic fluorite lattice of ceria
2 0 3 0 4 0 5 0 6 0 7 0 8 0 6 9 7 0 7 1 7 2
( 0 2 2 ) ( 1 1 0 )
( 0 0 2 ) ( 0 1 1 )
q /2 q /°
C e 0 . 6Z r0 . 4O 2
( 1 1 2 )
( 0 2 0 ) ( 1 2 1 )
( 0 1 3 )
( 0 0 4 )( 2 2 0 ) ( 1 2 3 )
( 0 3 1 ) ( 0 0 4 ) ( 1 1 1 )
( 0 0 2 )
Intensity (indiv. scaled)
C e 0 . 8Z r0 . 2O 2
( 0 2 2 ) ( 1 1 3 )
( 2 2 2 ) ( 1 3 3 )
( 1 3 3 ) ( 0 0 4 ) ( 2 2 2 )
( 1 1 3 ) ( 0 2 2 ) ( 0 0 2 )
C e O 2
( 1 1 1 )
P 4 2/ n m c
q /2 q /°
( 0 0 4 ) ( 2 2 0 )
( 0 0 4 ) F m 3 m
( 0 0 4 )
F m 3 m
Figure 4.1: Comparison of theθ/2θXRD scans of the 10CZO, 8CZO and 6CZO target. The intensity is scaled individually for every scan. The left graphs show the x-ray diffraction pattern as measured between20
◦
and 80
◦ with the respective reflex assignments based on the space group for the pattern given in the right graphs. In the right graphs the region between69
◦ and72
◦ are shown in detail. The ordinate is individually scaled for each graph. The x-axis scale is the same for each column of graphs in this figure.
becomes critically strained with an increasing concentration of smaller zirconium cations.
The ionic radius of a cerium cation isr(Ce
4+
) = 0.97Å whereas the radius of the zirconium cation is r(Zr
4+
) = 0.84Å.148 The reflexes shift towards higher diffraction angles, showing a decrease in the lattice constant from 10CZO to 8CZO. The X-ray diffraction pattern in figure 4.1 confirm this. Once the critical strain is reached, the cubic phase is not stable anymore and a tetragonal phase is formed. As the selection rules for a tetragonally strained cubic lattice change, the cubic reflexes split into two tetragonal ones. The larger the diffrac-tion angle, the larger is the splitting on the θ/2θ- axis. This can be seen in the right detail graphs in figure 4.1 for the cubic (004) reflex, splitting into two reflexes for the tetragonal 6CZO target. As the tetragonal distortion is only small for the CeO2 - ZrO2 system, the splitting is not visible for all reflexes. Especially for low order reflexes only a small asymmetry of the reflexes can be detected by x-ray diffraction. Thus, it cannot be excluded that a small amount of tetragonal material also exists in the 8CZO target material. But within this study and the sensitivity of the used methods no evidence for that was found.
The XRD analysis confirms the expected formation of a cubic or tetragonal single phase
Figure 4.2: SEM images (before the first use) of the targets used in the PLD process for this study. The 10CZO target has grain sizes that are in the range of that of the mother powder. The 6CZO surface shows evidence (black domains) for a second phase, like reported by Mamontov et al..92 The dark features on the 8CZO target surface are due to surface contaminants.
material as reported in the literature before. Whether local inhomogenities with a different crystallographic phase exist in the targets containing zirconium, as found by Mamontov et al., cannot be excluded, as the necessary neutron diffraction experiments were not conducted.92 Nevertheless, some evidence for this theory was found in the surface of the 6CZO target.
In the SEM image of the surface small domains (dots) within the grains were found that showed a lower contrast, using the secondary electron detector (SE2) that is sensitive to secondary electrons, as well as high energy backscattered electrons. The localized lower contrast indicates a different material or crystallographic surface from which less secondary and/or backscattered electrons are produced. On the contrary, the features observed here are several orders of magnitude larger than those reported by Mamontov et al., who found domains of Ce0.4Zr0.6O2 with a size of only 25−30Å in a matrix of Ce0.7Zr0.3O2.92 Thus, the features in the 6CZO target surface could very well also be minor phase impurities or originate from local differences in the crystallographic orientation towards the detector of the 6CZO phase itself.
Except for the shift of the reflexes and the splitting, the x-ray diffraction pattern also exhibit differences in intensities between the different targets, although the experimental setup and conditions used were the same. Especially the intensity of the 10CZO diffraction pattern is rather small compared to those of 8CZO and 6CZO. This is evidence that the crystalline quality in the 10CZO target is not as high as that of the 8CZO and 6CZO target. This observation can be explained by taking a look at the SEM images in figure 4.2. Here it is nicely resolved that the average crystallite size of the material in the sintered 10CZO target is significantly smaller than that of the targets containing zirconia, being an indicator for less perfectly grown crystals. As the melting temperature decreases when adding zirconia to ceria, the sintering temperature also decreases. For pure ceria the oven temperature of 1500°C during the sintering process was not high enough, promoting the growth of large crystallites. The melting temperature of pure ceria is about 2477°C.149 As a result of this
Intensity (a.u.)
1200 900 600 300 0
Binding Energy (eV) O1s
Ce4d
Zr3d
Ce3d
=O KLL rNa KLL ªC 1s
<Mg KLL
¨Al 2p
Zr3p<ª
=
Zr3sr
8CZO Target ¨
Ce4p
Figure 4.3:The graphs shows a survey spectra of the 8CZO target used for thin film ablation via pulsed laser deposition.
the 10CZO target still reveals open pore volume after the sintering process, while the other two targets show a dense surface. Porous targets often lead to a higher droplet formation during the PLD process. As for this study the formation of droplets was rarely observed using the porous 10CZO target, no attempt to enhance the density of the target by higher temperatures or other preparation techniques was made. For the 8CZO and 6CZO targets the temperature of1500
◦
C was sufficient to promote the sintering process. All three targets were mechanically stable, which was essential for the use in the PLD process.
In order to determine the composition of the targets and check for impurities not visible in the XRD experiments, the EDS data was evaluated. For 8CZO the sum formula was determined to be Ce0.79Zr0.21O2, assuming a stoichiometric amount of oxygen per formula unit. Small concentration of Ca (0.4 at.%) and Al (0.2 at.%) contaminants were found, which is almost below the detection limit of this technology. For the 6CZO target the sum formula was determined to be Ce0.60Zr0.40O2. No contaminants were found. The same holds true for the 10CZO target. Its composition was the same as that of the mother powder and no contaminants were found.
In addition to the EDS analysis, XPS measurements were performed on the 8CZO target as described in the experimental section. A survey spectrum of these experiments is shown in figure 4.3. From integrating the area below the Ce4d, Zr3d and O1s region one can estimate the composition. It was taken into account that two different oxygen species were detected, shown by two peaks in the O1s region. By determining the area below each peak separately
and multiplying it with their respective relative sensitivity factors, as well as the other areas, the Ce/Zr atomic ratio was determined to be 4.0, which is exactly the expected value.
Except of the Ce, Zr and O signals expected for the 8CZO target, the survey spectrum also contains signals of contaminants, like Mg, C and Al. As the sample was not sputter cleaned before being transferred into the XPS instrument, the detected carbon contaminations were within the typical intensities to be expected. The Mg and Al signals are more interesting.
Al on the surface could very well originate from the corundum crucible used in the sintering process. Using the Al2p signal for quantification the area corresponds to5.1 at.%of aluminum on the surface of the 8CZO target. In addition 3.3 at.%Mg were detected. Nevertheless, it can be concluded that the targets prepared were suitable to be used for the PLD process.
The result of this process is going to be described in the next section.